Drivers of Rural Households’ Choices and Intensity of Sustainable Energy Sources for Cooking and Lighting in Ondo State, Nigeria

Abstract

Poverty reduction and the promotion of sustainable human development are fundamentally dependent on having access to modern energy services. Energy supplies that are dependable, reasonably priced, and sustainable are vital to modern societies. In achieving the sustainable development goals (SDG7) and access to clean energy supplies, this study, using cross-sectional data from 180 randomly sampled rural households, analyzed the key factors determining the choice and intensity of energy sources used for lighting and cooking in rural Nigeria. Both descriptive and inferential statistics (multivariate probit (MVP) and zero-truncated Poisson (ZTP models)) were employed for the analyses. The result showed that there is evidence of fuel stacking in their choice of cooking and lighting energy, and it increases with rising income levels but is more pronounced for lighting than cooking. The result also revealed that reliable access to clean energy (9% of sampled households for LPG and 23% of the households for grid electricity) is very low, as these households still rely on fuelwood (70%) for cooking, but the predominant usage of kerosene (39%) for lighting, as reported in the literature, has drastically changed to dry cell battery (51%). The results using a multivariate probit model to capture the multiple fuel usage phenomenon among rural households show that access to clean energy, improvement in rural poverty, usage of indoor kitchens, household size, and an increase in the education of household heads’ spouses significantly influence the use of clean energy in the rural areas. In the same vein, the result of the ZTP model showed that income, access to energy sources, and occupation of the household head were the drivers of the intensity of cooking and lighting energy sources. Thus, it is recommended that any policy interventions that are targeted at encouraging rural households to use clean energy should start by improving rural access to these clean energy sources, improving their poverty status while also increasing the level of education and awareness of rural women concerning the risks of using dirty energy sources.

1. Introduction

Ensuring access to affordable, reliable, sustainable, and modern energy for all has been identified as one of the key sustainable development goals for 2030 (SDG7). According to the International Energy Agency (IEA) and World Bank [1], the UN Sustainable Energy for All initiative posits “sustainable energy for all” to encompass three pillars, namely energy access, energy security, and energy efficiency. The main policy concern about the usage of household energy in rural areas of most developing countries has always been how to understand the key factors that determine energy choices of rural households in order to encourage them to switch to modern energy sources for cooking and lighting, as these two energy services are basic energy needs that cut across all rural households, and the choices these households make can impact their livelihoods, health conditions, and the environment.

Approximately three billion people still rely primarily on wood or other biomass energy for cooking and heating. This accounts for nearly 1.06 billion people who lack access to electricity [2]. The incomplete combustion of this traditional biomass fuel leads to air pollution, which is estimated to cause 4.3 million deaths annually worldwide [3]. The majority of these individuals are from sub-Saharan African countries (SSA). Most of them reside in developing countries, mostly in Asia and Africa. Nine hundred and five million people, or the majority of those without access to clean cooking fuels, reside in sub-Saharan Africa. Furthermore, only 45% of people in SSA, home to 860 million people worldwide without access to electricity, live there. Based on current projections, the IEA estimates that 740 million people—up from 620 million in 2030—will not have access to electricity by 2050, and 1.8 billion people will still not have access to clean fuels for cooking in the next 20 years [4].

About 115 million people in Nigeria still primarily rely on traditional biomass as their primary source of energy for cooking, but every year, close to 79,000 people, mostly from the country’s rural and marginalized areas, pass away from pollution caused by the inefficient combustion of traditional biomass [5]. About 99.8% of cooking and heating in rural Nigeria are still done with traditional biomass, despite strong evidence that direct combustion of biomass energy, without the use of improved stoves, only offers very little real energy value with attendant environmental and health concerns, such as deforestation, land degradation, soil erosion, and air pollution [6,7,8].

Even though Nigeria has abundant primary energy resources to meet its domestic energy needs [9], the nation nevertheless has the second-largest population without access to electricity in the world (85 million people, of which 64% reside in rural areas). Furthermore, it is claimed that these rural residents use a variety of energy sources to meet their nighttime lighting needs [10]. In rural Nigeria, the absence of dependable electricity access would mean that households would be unable to improve their standard of living by extending their children’s study hours after school, improving their home’s nighttime lighting, or engaging in other profitable or productive endeavors. Another type of deprivation that exists in rural areas is poverty, which is closely related to rural households’ inability to access modern energy sources and other necessary resources for a suitable and sustainable way of life.

The reliance on solid biomass fuels and kerosene as primary energy sources for cooking and lighting, respectively, is extreme in rural areas of Nigeria. It is impossible to ignore the impact that poverty has on Nigeria’s rural areas’ patterns of energy consumption. Poor households are the most vulnerable in any society worldwide, according to [11]. Nigerian rural poverty is estimated to be 44.9%, compared to the country’s average of 33.1%, based on data from the General Household Surveys (GHS) panel format, which was conducted in 2010/2011 and 2012/2013 [12]. Given that the majority of impoverished rural households—particularly those in rural areas—cook their food outdoors on three-stone stoves, it is highly likely that these households experience hardships with food preparation and water heating, particularly during the rainy season. According to Louw et al. [13], in order to lessen their exposure to any unreliability linked to a single energy source, households typically use a combination of energy sources for a given energy service. Fuel stacking is the practice of rural households using a variety of energy sources [14]. When households use multiple fuels for a specific energy service, and their fuel combination pattern demonstrates complementarity between dirty and clean energy sources, even in the face of a significant improvement in their income or welfare status, this is known as fuel stacking. The phenomenon of fuel stacking in rural households is increasingly well-established in the literature, particularly with regard to cooking and lighting energy sources [15,16,17].

Similarly, the low penetration of electricity in rural areas of Nigeria has been ascribed to various factors, such as the expensive, insufficient, and unstable infrastructure brought on by aging transmission lines [18]. This has led to the over-reliance of a greater percentage of rural dwellers in Nigeria on traditional fuels like firewood and biomass to meet their household needs [19]. Relying on traditional fuels prevents nations from meeting their sustainable development targets because of the inability to support contemporary economic activities, such as heavy industries, and they impede social development by making it more difficult for people to access modern health and education services [20].

Though a number of studies conducted in Nigeria have demonstrated that households use a variety of energy sources for lighting and cooking [21,22,23], none of these studies take these various energy options into account in their various models. Similarly, researchers like Abdul-Wakeel and Dasmani [24], Abdul-Wakeel et al. [25], and Azorliade et al. [26] have concentrated on the factors that influence household cooking fuel preferences in developing nations, such as Nigeria, with some of these researchers examining the distinction between clean and dirty energy. These studies’ flaw is that they overlooked the problem of energy choice and consumption expenditure. Others, like Abdul-Wakeel et al. [27] and Ofori et al. [28], make an effort to address this problem by examining the effects of the fuel choice used in homes on human health, with a lot of their research concentrating on particular medical conditions. For instance, Ofori et al. [28] investigated the relationship between blood pressure and the use of dirty fuels in homes among women in southern Nigeria. To the best of our knowledge, this is the only study in Nigeria that models rural households’ choices of energy sources for lighting and cooking independently, analyzing the factors that influence these choices and providing a better understanding of the factors that influence the choices made. With regard to the intensity of energy sources for lighting and cooking in Nigeria, this study is helpful in developing effective policy interventions to promote the use of contemporary energy sources for these distinct energy services. The contribution of this study is premised on the comprehensive empirical analyses of the factors influencing rural households’ energy decisions in light of increased access to modern fuels and renewable energy sources relative to conventional energy sources while accounting for context-specific factors that influence household energy choices and intensity.

2. Materials and Methods

2.1. Study Area and Source of Data

The study was conducted in Ondo State, Nigeria. Ondo is one of the States in the southwestern part of Nigeria, where most of the rural areas have limited access to electricity and other modern energy. Primary data were used for this study, and the data were collected with the aid of a well-structured questionnaire. A multistage sampling technique was used to select respondents for this study. The first stage was a purposive selection of Okitipupa and Owo Agricultural Development Programme (ADP) zones out of the three existing ADP zones in the State. Each zone comprises six local government areas (LGAs). In the second stage, two LGAs were selected from each zone using a simple random sampling technique. In the third stage, simple random sampling was also used to select three villages from each of the LGAs. The final stage was the random selection of fifteen households each from the selected villages/communities to make a total sample size of one hundred and eighty (180) respondents. Detailed information on the choice, access, usage, and purchase of household energy sources for cooking and lighting were collected alongside key demographic and socioeconomic characteristics from each of the rural households.

2.2. Empirical Model

A discrete choice model is the multivariate probit model. Similar to the seemingly unrelated regressions model, it is a multiple-equation extension of the probit model that permits the disturbance terms to be correlated [29]. This model is very helpful in managing fuel stacking because it permits the error terms to be freely correlated, in addition to enabling the simultaneous analysis of these various fuel options. The factors influencing rural households’ decisions about cooking energy sources were estimated using a system of 4-equation multivariate probit model, in accordance with [30], while a system of 5-equation multivariate probit model was employed for their lighting energy source choices.

$$y_{im}^{*}={\beta }_m^{\prime }{X}_{im}+{\varphi }_{im,} m=1,...,M$$

$$y_{im}=1 \mathrm{if} y_{im}^{*}>0\mathrm{and}0\mathrm{otherwise}$$

${ϵ}_{im}, m=1, \dots , M$ are error terms spread as a variance–covariance matrix and multivariate normal, each with a mean of zero. V, where V has values of 1 on the leading diagonal and correlations ${\rho }_{jk}={\rho }_{kj}$ as off-diagonal elements. $X_{im}=$ vector of the explanatory variables and ${{\beta }_m}^{\prime }$ = vector with to-be-estimated unknown coefficients. The outcome variables for rural households’ choices of cooking energy are fuelwood (FWD), charcoal (CHL), kerosene (KRC), and bottled liquefied petroleum gas (LPG), while those for lighting are kerosene (KRL), dry cell battery (DCB), rechargeable battery (RGB), petrol (PMS), and grid electricity (GEL).

Thus, we assume that there is independence in the error terms of equations for both models (cooking and lighting) in order to test for the absence of correlation among the error terms. That is, all cross-equation correlation coefficients for each model are jointly equal to zero (for example, rho21 = rho31 = rho41 = rho32 = rho42 = rho43 = 0). The likelihood ratio test is among the appropriate tests used to test the null hypothesis of no correlation of error terms across a multivariate probit model [30]. Without significant statistical evidence to reject the null hypothesis, it means that the choices of rural households concerning a particular energy service are not jointly made, and in that case, it would be better to estimate these models separately. However, if this null hypothesis is statistically rejected, estimating the models separately will result in inefficient and biased estimates.

In estimating the drivers of the intensity of level of energy sources (cooking and lightning) among the sampled rural households in Ondo State, a zero-truncated Poisson regression model was employed, as used by Ogundari [31]. The outcome variable is specified as a count data. Empirical studies with cross-sectional data are hypothesized to have $n$ explanatory observations, such that the $i^{th}$ observation denoted as $(z_i,x_i)$. $z_i$ in the current study is the energy source for cooking and lighting, which is the number of energy sources used by the rural households. $x_i$ is the vector of linearly dependent variables that are likely to determine the dependent variable $z_i$. Independent variables $x_i$ include household socioeconomic and demographic variables. Assuming randomness within a specified time interval, the household energy sources for cooking and lighting have a Poisson distribution with a probability density function mathematically defined as:

$$f(z_i={{\rm Y}}_i|x_i)\frac{e^{-{\varpi }_i}{\varpi }_i^{{\rm Y}}}{{{\rm Y}}_i!}, {{\rm Y}}_{i=1, 2, 3\dots 10}$$

${{\rm Y}}_i$ is the random variable’s actual value, which has a mean $z_i$ and variance ${\varpi }_i$. It is hypothesized that $z_i$ is strictly positive $(z_i>0)$, which suggests a zero-truncated outcome variable. Explicitly stated, the log-linear version of the estimated model can be specified as:

$$E[z_i|x_i]={\varpi }_i=\exp (x_i^{\prime }{\theta }_i)=\exp (x_{1i}^{\prime }{\theta }_i)\exp (x_{2i}^{\prime }{\theta }_2)...\exp (x_{ji}^{\prime }{\theta }_j)$$

The additive specification of the above equation is:

$$E[z_i|x_i]=x_i^{\prime }\beta ={\displaystyle {\sum }_{j=1}^k}xji{\beta }_j$$

Based on the property of Poisson, ${\varpi }_i={\lambda }_i$ is predicated on equidispersion. Cameron and Trivedi [32] claim that the qualitative impact of the Poisson assumption of equidispersion failing is comparable to that of the homoscedasticity assumption failing in a linear regression model.

3. Results

3.1. Descriptive Statistics of the Respondents

The descriptive statistics of the variables used in the multivariate probit models are presented in

Table 1. Descriptive statistics of variables for the models.

Variables	Description	Mean	Std Dev.
Dependent variables for cooking model
Fuelwood (Y1)	Dummy = 1 if household used fuelwood for cooking, 0 otherwise	0.7000	0.4595
Charcoal (Y2)	Dummy = 1 if household used charcoal for cooking, 0 otherwise	0.2111	0.4092
Kerosene (Y3)	Dummy = 1 if household used kerosene for cooking, 0 otherwise	0.4833	0.5011
LPG (Y4)	Dummy = 1 if household used LPG for cooking, 0 otherwise	0.1111	0.3151
Dependent variables for lighting model
Kerosene	Dummy = 1 if household used kerosene for lighting, 0 otherwise	0.3944	0.4900
Dry cell battery	Dummy = 1 if household used dry cell battery for lighting, 0 otherwise	0.5111	0.5012
Rechargeable batter	Dummy = 1 if household used rechargeable battery for lighting, 0 otherwise	0.2833	0.4518
Petrol	Dummy = 1 if household used petrol for lighting, 0 otherwise	0.2944	0.4570
Grid electricity	Dummy = 1 if household used grid electricity for lighting, 0 otherwise	0.3500	0.4783
Intensity of lighting	Number of energy sources used for lighting	1.8333	0.7436
Intensity of cooking	Number of energy sources used for cooking	1.5052	0.6023
Explanatory variables
Age	Age of household head in years	48.461	11.2827
Household size	Number of household members	5.2611	1.9700
Elec_Access	Dummy = 1 if household had reliable access to grid electricity, 0 otherwise	0.2277	0.4205
LPG_Access	Dummy = 1 if household had reliable access to LPG, 0 otherwise	0.0944	0.2932
Own_Gen	Dummy = 1 if household owns an electricity generator, 0 otherwise	0.3833	0.4875
Kitchen type	Dummy = 1 if household cooks in indoor kitchen, 0 otherwise	0.2388	0.4275
Education	Years of formal education completed by household head	9.7667	5.6033
Edu_Spouse	Years of formal education completed by household head’s spouse	6.7667	5.8118
Food expenditure	Household expenditure on food in the last 30 days (in Naira)	13448.1	6147.3
Poverty status	Dummy = 1 if household per capita monthly expenditure is above NGN 4,405.19, 0 otherwise	0.6833	0.4664
Farmer	Dummy = 1 if household head main occupation is farming, 0 otherwise	0.4333	0.4969
Artisan	Dummy = 1 if household head main occupation is artisan, 0 otherwise	0.2944	0.4570
Trader	Dummy = 1 if household head main occupation is trading, 0 otherwise	0.0388	0.1938
Civil servant	Dummy = 1 if household head main occupation is in the civil service, 0 otherwise	0.2333	0.4241
Own_SATtv	Dummy = 1 if household owns a satellite television in their house, 0 otherwise	0.3500	0.4783
Owner	Dummy = 1 if the dwelling space is owned by the household, 0 otherwise	0.4778	0.5008

. About 70 percent of the sampled households used fuelwood for cooking, while 21, 48, and 11 percent of them used charcoal, kerosene, and LPG for cooking, respectively. This indicates that rural households are still heavily dependent on dirty energy sources for their cooking. On the other hand, 39 percent of these households used kerosene for lighting, while 51, 28, 29, and 35 percent of them used dry cell batteries, rechargeable batteries, petrol, and grid electricity for lighting, respectively. The data also show the average age of household heads to be 48 years, with an average household having five members. On reliable access to modern energy in rural areas, twenty-two percent of the households sampled had access to grid electricity, while only nine percent reported having access to bottled liquefied petroleum gas (LPG). Also, 38 percent of the households reported owning an electricity generator, and only 23 percent cook their food in an indoor kitchen.

Rural households’ lighting and cooking energy consumption (choice) pattern

More so, rural households’ cooking energy consumption (choice) pattern is depicted according to their respective income quintiles (Figure 1).

Furthermore, rural households’ lighting energy consumption (choice) pattern is depicted according to their respective income quintiles (Figure 2). The figure shows that rural households from the lowest income quintile to the highest income quintile use petrol for lighting, and the frequency of use increases with income level.

Distribution of cooking and lighting energy choices by income quintiles

The distribution of rural households based on their different choices of cooking energy across income quintiles (

Table 2. Distribution of cooking and lighting energy choices by income quintiles.

	Distribution of Cooking Energy Choices by Income Quintiles (Pooled Data)
Variables	Income Quintile 1 (n = 36)	Income Quintile 2 (n = 34)	Income Quintile 3 (n = 38)	Income Quintile 4 (n = 36)	Income Quintile 5 (n = 36)	All Respondents (n = 180)
Fuelwood	32 [74.42]	25 [51.02]	27 [54.00]	26 [44.83]	16 [22.54]	126
Charcoal	5 [11.63]	7 [14.29]	6 [12.00]	8 [13.79]	12 [16.90]	38
Kerosene	6 [13.95]	17 [34.69]	16 [32.00]	22 [37.93]	26 [36.62]	87
LPG	0 [0.00]	0 [0.00]	1 [2.00]	2 [3.45]	17 [23.94]	20
Total	43	49	50	58	71
	Distribution of cooking energy choices by income quintiles (AEZ 1)
Fuelwood	20 [68.97]	13 [46.43]	14 [58.33]	11 [61.11]	9 [21.95]	67
Charcoal	4 [13.79]	6 [21.43]	2 [8.33]	1 [5.56]	15 [36.59]	28
Kerosene	5 [17.24]	9 [32.14]	8 [33.33]	6 [33.33]	10 [24.39]	38
LPG	0 [0.00]	0 [0.00]	0 [0.00]	0 [0.00]	7 [17.07]	7
Total	29	28	24	18	41
	Distribution of cooking energy choices by income quintiles (AEZ 2)
Fuelwood	12 [85.71]	12 [57.14]	13 [50.00]	15 [37.5]	7 [17.50]	59
Charcoal	1 [7.14]	1 [4.76]	4 [15.38]	7 [17.5]	7 [17.50]	20
Kerosene	1 [7.14]	8 [38.10]	8 [30.77]	16 [40.00]	16 [40.00]	49
LPG	0 [0.00]	0 [0.00]	1 [3.85]	2 [5.00]	10 [25.00]	13
Total	14	21	26	40	40

) shows that most of these households cook with fuelwood, with the exception of households in the highest income quintile ((36.62%) and (23.94%).) that cook mostly with kerosene.

The implication of the results in Table 2 shows that the improvement in the economic or wealth status of rural households will increase their usage of modern and clean energy for cooking. The results in the lower part of Table 2 enable us to take a closer look at our findings across different agroecological zones (AEZ 1 and 2) in our study area. The majority of households in AEZ1, within the income quintiles 1–4, use fuelwood for cooking, while most of the households in income quintile 5 cook with charcoal (36.59%) and kerosene (24.39%). Also, the households in income quintiles 1–3 located in AEZ 2 show similar patterns with their counterparts in AEZ 1. However, most households in income quintile 4 cook with fuelwood (37.5%) and kerosene (40.00%), while most households in income quintile 5 cook with kerosene (40.00%) and LPG (25.00%).

The consumption of lighting energy among rural households (

Table 3. Distribution of lighting energy choices by income quintiles.

	Distribution of Lighting Energy Choices by Income Quintiles (Pooled Data)
Variables	Income Quintile 1 (n = 36)	Income Quintile 2 (n = 34)	Income Quintile 3 (n = 38)	Income Quintile 4 (n = 36)	Income Quintile 5 (n = 36)	All Respondents (n = 180)
Kerosene	17 [36.17]	11 [18.97]	22 33.85]	13 [17.81]	8 [9.20]	71
Dry cell battery	16 [34.04]	27 [46.55]	12 18.46]	17 [23.29]	20 [22.99]	92
Petrol	1 [2.13]	4 [6.89]	8 [12.31]	13 [17.81]	27 [31.03]	53
Grid electricity	7 [14.89]	12 [20.69]	11 [16.92]	18 [24.66]	15 [17.24]	63
Rechargeable battery	6 [12.77]	4 [6.90]	12 [18.46]	12 [16.44]	17 [19.54]	51
Total	47	58	65	73	87
	Distribution of lighting energy choices by income quintiles (AEZ 1)
Kerosene	13 [50.00]	7 [24.14]	13 [41.93]	7 [28.00]	3 [8.82]	43
Dry cell battery	10 [38.46]	17 [58.62]	5 [16.13]	6 [24.00]	13 [38.24]	51
Petrol	1 [3.85]	3 [10.34]	7 [22.58]	8 [32.00]	14 [41.18]	33
Grid electricity	0 [0.00]	0 [0.00]	0 [0.00]	0 [0.00]	0 [0.00]	0
Rechargeable battery	2 [7.69]	2 [6.90]	6 [19.36]	4 [16.00]	4 [11.76]	18
Total	26	29	31	25	34
	Distribution of lighting energy choices by income quintiles (AEZ 2)
Kerosene	4 [19.05]	4 [13.79]	9 [26.47]	6 [12.50]	5 [9.43]	28
Dry cell battery	6 [28.57]	10 [34.48]	7 [20.59]	11 [22.92]	7 [13.21]	41
Petrol	0 [0.00]	1 [3.45]	1 [2.94]	5 [10.41]	13 [24.53]	20
Grid electricity	7 [33.33]	12 [41.38]	11 [32.35]	18 [37.50]	15 [28.30]	63
Rechargeable battery	4 [19.05]	2 [6.90]	6 [17.65]	8 [16.67]	13 [24.53]	33
Total	21	29	34	48	53

) shows that there is stacking of lighting energy sources as income progresses. Dry cell batteries are mostly used for lighting in rural households. The lowest income quintile households mostly kerosene (36.17%) and dry cell batteries (34.04%) for lighting, while the highest income quintile households use petrol (31.03%), dry cell batteries (22.99%), and rechargeable batteries (19.54%) for lighting up their houses at night. The results in the lower part of Table 3 show that all the households in AEZ 1 did not use grid electricity. This is because AEZ 1 was disconnected from the national grid by the electricity distribution company serving the area. While grid electricity was the highest source of lighting energy for rural households in AEZ 2, most rural households in AEZ 1 relied on dry cell batteries to light up their homes and environments.

3.2. Multivariate Probit Model on Choices of Cooking Energy among Rural Households

This section discusses the results from the multivariate probit model on choices of cooking energy among rural households. The likelihood ratio test (chi2(6) = 15.9343, p < 0.01) of the independence of the error terms of the different cooking energy choice equations is highly rejected (

Table 4. Correlation coefficients of households’ cooking energy choices.

Cooking Energy Choices	Correlation Coefficient	Standard Error
Charcoal and Fuelwood	−0.2507 *	0.1439
Kerosene and Fuelwood	−0.4552 ***	0.1409
LPG and Fuelwood	−0.1440	0.2906
Kerosene and Charcoal	−0.1692	0.1482
LPG and Charcoal	−0.3699 *	0.2255
LPG and Kerosene	0.1347	0.2211
Prob > chi2 Chi2(10)	0.0141 15.9343
Likelihood ratio test of rho21 = rho31 = rho41 = rho32 = rho42 = rho43 = 0

Note: *, and *** indicate statistical significance at the 10% and 1% alpha levels, respectively.

). We thus adopt the alternative hypothesis of the mutual interdependence among the choice of cooking energy. The result thus supports the use of a multivariate probit model to analyze the determinants of household fuel choice when there is evidence of fuel stacking.

Table 4 presents the pairwise correlation coefficients showing the relationship between various cooking energy source choices made by households.

The results of parameters estimated from the multivariate probit model on household cooking energy choice are presented in

Table 5. Parameter estimates from multivariate probit for household cooking energy choices.

Variables	Fuelwood	Charcoal	Kerosene	LPG
Age	0.0126 (0.0153)	0.0219 * (0.0127)	−0.0071 (0.0113)	0.0187 (0.0212)
Household Size	0.3823 *** (0.1066)	−0.1441 * (0.8280)	−0.1154 * (0.0675)	−0.4622 ** (0.2018)
Access to LPG	−1.3807 * (0.7445)	−0.0895 (0.4887)	−0.0065 (0.4702)	2.9051 *** (0.7185)
Type of Kitchen	−1.0202 *** (0.3141)	0.2925 (0.3196)	0.6162 ** (0.3188)	1.2376 *** (0.4472)
HH’s Education	−0.0764 ** (0.0390)	0.0270 (0.0374)	−0.0249 (0.0295)	−0.1142 (0.0792)
Spouse’s Education	−0.0662 ** (0.0302)	0.0546 ** (0.0278)	0.0943 *** (0.0251)	0.2025 *** (0.0783)
Monthly Food Expenditure	−3.40 × 10−5 (2.37 × 10−5)	2.82 × 10−5 (2.52× 10−5)	1.78 × 10−5 (2.07 × 10−5)	7.07 × 10−5 ** (3.60 × 10−5)
Poverty Status (poor = 0)	−0.8391 ** (0.4159)	0.2026 (0.3329)	0.5343 ** (0.2774)	0.8186 * (0.4703)
HH’s Main Occupation
Artisans	−1.7013 *** (0.4301)	0.8714 ** (0.3715)	0.5448 ** (0.2717)	0.3417 (0.8137)
Traders	−1.1340 ** (0.5606)	1.4127 ** (0.6145)	0.8205 * (0.4674)	−3.7122 *** (0.8897)
Civil Servants	−1.0121 ** (0.4328)	1.0026 ** (0.4257)	0.7861 ** (0.3521)	0.6316 (0.7419)
_Constants	2.2891 ** (0.9993)	−2.4461 ** (0.9002)	−0.6067 (0.6504)	−3.7666 *** (1.4464)
Prob > chi2 =	0.0141 **
chi2(10)	15.9343
Likelihood ratio test of rho21 = rho31 = rho41 = rho32 = rho42 = rho43 = 0:

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% alpha levels, respectively.

. We found out that households with older household heads are more likely to use charcoal as their cooking energy choice. This is in support of results from Gebreegziabher et al. [33] that older household heads in Ethiopia are more likely to consume charcoal for cooking. The result of the education status shows that household heads who are more educated are less likely to use fuelwood as their choice of cooking energy in rural areas. An increase in a household’s monthly expenditure on food will increase the likelihood of using LPG as the choice of cooking energy. This is in line with the results of Ogwumike et al. [22] that an increase in per capita expenditure increases the probability that the household will use LPG for cooking.

3.3. Multivariate Probit Model on Choices of Lighting Energy among Rural Households

This section discusses the results from the multivariate probit model on choices of lighting energy among rural households.

Table 6. Correlation coefficients of household lighting energy choices.

Lighting Energy Choices	Correlation Coefficients	Standard Errors
Dry cell battery and Kerosene	−0.7386 ***	0.0895
Petrol and Kerosene	−0.5278 ***	0.1083
Grid electricity and Kerosene	0.0757	0.1565
Rechargeable battery and Kerosene	−0.3196 ***	0.1156
Petrol and Dry cell battery	0.0469	0.1729
Grid electricity and Dry cell battery	−0.0337	0.1602
Rechargeable battery and Dry cell battery	−0.2009	0.1317
Grid electricity and Petrol	−0.3799	0.2691
Rechargeable battery and Petrol	0.3295 **	0.1508
Rechargeable battery and Grid electricity	0.1144	0.1748
Prob > chi2	0.0000 ***
Chi2(10)	74. 2261
Likelihood ratio test of rho21 = rho31 = rho41 = rho51 = rho32 = rho42 = rho52 = rho43 = rho53 = rho54 = 0:

Note: **, and *** indicate statistical significance at the 5%, and 1% alpha levels, respectively.

presents the pairwise correlation coefficients showing the relationship between various lighting energy choices made by households. A negative correlation between two lighting energy sources indicates substitutability, while a positive correlation means complementarity.

Table 7. Parameter estimates from multivariate probit for household lighting energy choices.

Variables	Kerosene	Battery	Petrol	Grid	R-Battery
Age	0.0108 (0.0127)	−0.0182 (0.0119)	−0.0158 (0.0152)	0.0259 * (0.0155)	0.0134 (0.0134)
Household size	−0.0319 (0.0692)	0.0385 (0.0689)	0.2607 *** (0.0950)	0.1823 ** (0.0886)	−0.0583 (0.0667)
Access to grid electricity	−1.2739 *** (0.3255)	−0.4544 * (0.2622)	−0.3234 (0.3245)	3.4641 *** (0.5406)	0.9564 *** (0.3286)
Ownership of generator set	−0.3294 (0.2621)	−0.2402 (0.2683)	2.4222 *** (0.3473)	−0.8766 ** (0.4010)	0.7305 ** (0.3021)
Access to LPG	−0.6599 (0.4511)	−0.6131 (0.4127)	1.7045 *** (0.5297)	0.1841 (0.5466)	−0.9052 * (0.5323)
Type of kitchen	−0.1374 (0.2843)	0.3537 (0.3142)	−0.5338 (0.3712)	0.0213 (0.4028)	0.1165 (0.3341)
HH’s education	−0.0086 (0.0246)	0.0015 (0.0243)	0.1021 *** (0.0332)	−0.0529 * (0.0313)	−0.0012 (0.0289)
Spouse’s education	0.0010 (0.0242)	0.0104 (0.0222)	−0.0283 (0.0251)	−0.0302 (0.0294)	0.0191 (0.0264)
Monthly food Expenditure	4.22 × 10−5 * (2.25 × 10−5)	−5.99 × 10−7 (2.20 × 10−5)	6.44 × 10−6 (2.77 × 10−5)	−4.80 × 10−5 (3.02 × 10−5)	−6.90 × 10−6 (2.41 × 10−5)
Poverty status (poor = 0)	−0.2069 (0.2659)	0.0078 (0.2465)	0.4921 (0.3232)	0.5901 ** (0.3058)	0.1578 (0.2851)
HH’s main occupation
Artisans	−0.6166 ** (0.2839)	0.1429 (0.2757)	0.4505 (0.3893)	−0.0781 (0.4232)	−0.6932 ** (0.3210)
Traders	0.2415 (0.4778)	0.1331 (0.5109)	1.0457 * (0.6216)	−1.5208 ** (0.7274)	−0.7638 (1.0602)
Civil servants	−1.0273 *** (0.3763)	0.7279 ** (0.3677)	0.5003 (0.4273)	−0.0211 (0.4974)	0.1083 (0.3684)
Satellite TV	0.2821 (0.2912)	−0.6907 ** (0.2930)	0.1847 (0.3979)	0.6487 * (0.3692)	0.6310 ** (0.2862)
Ownership of house	−0.1502 (0.2569)	0.2803 (0.2461)	0.4571 (0.3085)	−0.4900 (0.3182)	−0.2995 (0.2506)
_Constants	−0.1899 (0.7146)	0.6581 (0.6174)	−4.4530 *** (1.0873)	−3.0195 *** (0.8467)	−1.4946 ** (0.7348)
chi2(10)	74.2261
Prob > chi2	0.0000 ***
Observation	180
Log pseudolikelihood	−340.1692
Prob > χ2	0.000 ***
Wald χ2 (75)	450.40
Likelihood ratio test of rho21 = rho31 = rho41 = rho51 = rho32 = rho42 = rho52 = rho43 = rho53 = rho54 = 0

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% alpha levels, respectively.

presents the results of parameter estimates from the multivariate probit model on household lighting energy choice. An increase in the household head’s age will make the usage of grid electricity as the household’s choice of lighting energy source more likely.

3.4. The Drivers of Intensity of the Use of Both Cooking and Lighting Energy among Rural Households

The empirical result of drivers of the intensity of the use of cooking and lighting energy among rural households using the zero-truncated Poisson model is presented in

Table 8. Drivers of intensity of the use of both cooking and lighting energy among rural households using the zero-truncated Poisson model.

	Lighting Energy Sources for Lighting				Cooking Energy Sources for Lighting
	Coeff.	Std. Err.	dy/dx	Std.Err	Coeff.	Std.Err	dy/dx	Std.
Gender	0.269	0.250	0.322	0.283	0.135	0.305	0.103	0.226
Age	−0.002	0.010	−0.003	0.012	0.009	0.011	0.007	0.009
Farm size	−0.000	0.078	−0.001	0.098	0.097	0.105	0.076	0.082
Formal education	0.430	0.277	0.471	0.260	0.031	0.363	0.024	0.277
Household size	0.075	0.061	0.094	0.076	−0.126	0.082	−0.098	0.063
Marital status	0.160	0.156	0.201	0.195	0.073	0.176	0.057	0.137
LPG access	−0.119	0.065 *	−0.149	0.081 *	−0.079	0.083	−0.062	0.065
Total income	0.000	0.000 ***	0.000	0.000 ***	0.000	0.000 **	0.000	0.000
Occupation	−0.001	0.073	−0.002	0.092	0.201	0.100 **	0.157	0.076
_cons	−0.099	0.706			−0.861	0.878
Prob > chi2	0.000				0.000
LR chi2(11)	42.53				58.23
Pseudo R2	0.0978				0.1600
Log likelihood	−196.09				−152.84

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% alpha levels, respectively.

.

The coefficient of total income of the household head is positive and statistically significant in influencing the intensity of usage of energy sources for lighting and cooking in the study area. Households switch from using traditional fuels like wood to transitional fuels like kerosene and finally to modern fuels like electricity from the grid as the head of the household’s income rises [22]. Access to affordable and sustainable energy services is a prerequisite for achieving the internationally recognized goal of halving the percentage of the population living on less than USD 1 per day by 2015.

4. Discussion

4.1. Descriptive Statistics of the Respondents

Contrary to evidence from several studies in Nigeria [22] that the majority of households use kerosene for lighting, the possible explanation for this could be either the supply of kerosene to the rural areas is becoming more unreliable and expensive or rural households are becoming more self-aware of the hazards related to the use of kerosene for lighting in their households. More so, an average household head completed 10 years of formal education, while the average years of formal education completed by the household head’s spouse was 7 years. The number of years of formal education completed by the household head’s spouse could be an important indicator of her empowerment in the house, and it can also mean that the spouse is more enlightened about the danger of using dirty energy sources. An average household for this study spent NGN 13,448 monthly on food. The poverty status shows that 68 percent of households live above the study poverty line. In terms of the main occupation, 43.33 percent of the household heads are farmers, while 29.44, 3.88, and 23.33 percent are artisans, traders, and civil servants, respectively.

4.2. Rural Households’ Lighting and Cooking Energy Consumption (Choice) Pattern

Figure 1 shows that rural households from the middle-income quintile to the higher-income quintile use modern energy (LPG) for cooking, and the frequency of use increases with income level. It also shows that there is evidence of fuel stacking across different income quintiles, but the level of fuel stacking increases with high-income level households. This implies that low-income rural households are more vulnerable to the unreliable supply of a single energy source (especially modern energy) than their high-income neighbors who have the incentive to use more than one source of energy for cooking. This is corroborated by Louw et al. [13] that households mostly use multiple energy sources to forestall their vulnerability to the failure of a single fuel that is mostly unreliable. There is evidence of fuel stacking for lighting energy across different income quintiles in the study area, and the evidence becomes stronger with an increase in the income level of rural households. This implies that the higher the income level of households in rural areas, the higher the likelihood of fuel stacking for lighting. A comparison of Figure 1 and Figure 2 shows that there is more stacking in lighting energy sources than cooking energy sources among rural households and across different income levels. This implies that households may be more vulnerable in terms of their lighting energy choice than their cooking energy choice. The possible explanation for this is that the supply of fuelwood (one of the energy sources for cooking) is within the domain of rural households, while almost all the energy sources for lighting are supplied outside their domain.

4.3. Distribution of Cooking and Lighting Energy Choices by Income Quintiles

In comparison, the results (AEZ 1 and AEZ 2) show that high-income rural households in AEZ 2 use LPG for cooking more than their counterparts in AEZ 1. While there is a reduction in the usage of fuelwood among households in AEZ 1 as their income increases, the usage of fuelwood is roughly the same among households in AEZ 2, except for households in quintile 5.

4.4. Multivariate Probit Model on Choices of Cooking Energy among Rural Households

A negative correlation between two cooking energy sources indicates substitutability, while a positive correlation means complementarity. Interestingly, the result shows a negative correlation between LPG (a clean cooking energy source) and dirty energy sources (fuelwood and charcoal). This implies that improvement in poverty status and reliable access to LPG in rural households will probably shift to the use of clean energy sources for cooking. Also, there is a positive correlation between LPG and kerosene, indicating that households that use this clean cooking energy source complement it with kerosene. In general, Table 4 supports the idea that households typically rely on multiple energy sources. For instance, a household might rely on cooking with both kerosene and LPG. Because energy sources can coexist in a single household, we can estimate household preferences for various cooking energy sources by using a multivariate probit model.

Households with large member sizes are more likely to use fuelwood but less likely to use charcoal, kerosene, and LPG as their choice of cooking energy in rural areas. Findings from the studies of Pandey and Chaubal [34] and Özcan et al. [35] are in support that larger households prefer dirty fuels over clean fuels in most developing countries. Households that have reliable access to LPG are less likely to use fuelwood but more likely to use LPG as their choice of cooking energy. This implies that rural households that find LPG accessible, unlike those that find it non-accessible, are more likely to use LPG as their choice of cooking energy. This is in line with Mensah and Adu [36], who posited that ensuring households have access to LPG for cooking can drive the move towards cleaner fuels.

This result shows that educated household heads will have a better understanding and awareness of the risks associated with cooking with fuelwood. This supports the findings of Alem et al. [37] and Ogwumike et al. [22] that households with more educated household heads are less likely to use fuelwood for cooking. More so, an increase in the years of education attained by the household head’s spouse will reduce the likelihood of using fuelwood as the choice of cooking energy, but it will also make the use of charcoal, kerosene, and LPG more likely. This is because the household head’s spouse with a higher education level has better empowerment within the house as she is more likely to be engaged in more productive activities that will increase the opportunity cost of using fuelwood for cooking. Baiyegunhi and Hassan [38] observe that a higher education level induces households to shift away from fuelwood towards the use of kerosene and LPG.

An improvement in the poverty status of rural households will not only reduce the likelihood of using fuelwood as their choice of cooking energy, but it will also increase the likelihood of using kerosene and LPG for cooking. This is in line with the findings of Mensah and Adu [36] that moving from extremely poor to a non-poor welfare status reduces the probability of a household using crop residue and firewood while increasing the probability of switching to relatively cleaner fuels like charcoal, kerosene, and LPG. Household heads who are artisans are less likely to use fuelwood, but they are more likely to use charcoal and kerosene as their choice of cooking energy compared to household heads who are farmers. Also, household heads who are traders are less likely to use both fuelwood and LPG, but they are more likely to use charcoal and kerosene as their choice of cooking energy when compared to household heads who are farmers. More so, when compared to household heads who are farmers, civil servant household heads are less likely to use fuelwood, but they are more likely to use charcoal and kerosene as their choice of cooking energy in the study area.

4.5. Multivariate Probit Model on Choices of Lighting Energy among Rural Households

The likelihood ratio test (chi2(10) = 74.2261, p < 0.0001) of the independence of the error terms of the different lighting energy choice equations is highly rejected (Table 6). We thus adopt the alternative hypothesis of the mutual interdependence among the choice of lighting energy. The result thus supports the use of a multivariate probit model to analyze the determinants of household fuel choice when there is evidence of fuel stacking.

Interestingly, the result shows a negative and significant correlation between dry cell battery and kerosene, petrol and kerosene, and rechargeable battery and kerosene. This shows that there is a high likelihood that rural households will shift away from kerosene as their lighting energy choice if provided with cleaner alternatives. Also, there is a positive and significant correlation between rechargeable batteries and petrol, indicating that households that use petrol as their lighting energy source complement it with a rechargeable battery. There is also a positive correlation between rechargeable batteries and grid electricity, but it is not statistically significant. This could be a result of a low level of access to grid electricity in rural areas. Table 6 generally confirms that households usually depend on more than a single source of energy for lighting.

Households with large member sizes are more likely to use both petrol and grid electricity as their choice of lighting energy. This is also supported by Giri and Goswami [39]. Households that have reliable access to grid electricity are less likely to use kerosene and dry cell batteries but more likely to use grid electricity and rechargeable batteries as their choice of lighting energy. This implies that providing rural households with a reliable supply or access to electricity will not only enhance the usage of grid electricity and rechargeable batteries but will also significantly discourage the use of both kerosene and dry cell batteries for lighting. This finding is also supported by Lay et al. [40]. Households that own electricity-generating sets are more likely to use petrol and rechargeable batteries as their choice of lighting energy, but they are less likely to use grid electricity. The reason for this could be that the majority of households that own electricity-generating sets do not have reliable access to electricity, as the electrification rate in rural Nigeria is only 35 percent [10].

Households that have reliable access to LPG are less likely to use rechargeable batteries but more likely to use petrol as their choice of lighting energy. This could be that households that reported to have reliable access are high-income level households and, as such, they could afford to use petrol for lighting in the absence of a reliable supply of electricity. Household heads who are more educated are less likely to use grid electricity but more likely to use petrol as their choice of lighting energy in rural areas. An increase in the household’s monthly expenditure on food will increase the likelihood of using kerosene as the choice of lighting energy. An improvement in the poverty status of rural households will increase the likelihood of using grid electricity as their choice of lighting energy source. This implies that alleviating rural poverty will significantly increase the usage of clean modern energy sources for lighting while discouraging the use of dirty energy sources. Household heads who are artisans are less likely to use both kerosene and rechargeable batteries as their choice of lighting energy compared to household heads who are farmers. Also, household heads who are traders are less likely to use both grid electricity, but they are more likely to use petrol as their choice of lighting energy when compared to household heads who are farmers. More so, when compared to household heads who are farmers, civil servant household heads are less likely to use kerosene, but they are more likely to use dry cell batteries as their choice of lighting energy in the study area. Furthermore, households who have satellite television are more likely to use grid electricity and rechargeable batteries as their choices of lighting energy sources, but they are less likely to use dry cell batteries.

4.6. The Drivers of Intensity of the Use of Both Cooking and Lighting Energy among Rural Households

The results imply that as the total income of the respondents increases, the propensity to increase the number of energy sources also increases. The results could mean that households are switching from traditional solid cooking fuels to more modern and efficient clean energy sources like LPG, electricity, and solar energy as people’s incomes rise and other socioeconomic characteristics change. This outcome aligns with the energy-ladder theory. According to the hypothesis, a household’s energy sources are significantly influenced by its income level [41]. This theory supports the idea behind the economic theory of the consumer, which states that when income increases, consumers not only demand more of a good but also shift their consumption habits to prioritize higher-quality products.

Modern clean fuels are generally thought to be more efficient, comfortable, and user-friendly than traditional or transitional fuels. Similarly, studies by Ravindra et al. [42] and Kapsalyamova et al. [43] suggest that household head adoption of energy sources for cooking and lighting in the study area is significantly influenced by income and affordability, even in light of the recent increase in the price of clean cooking fuel.

According to this study, the respondent’s intensity of use of energy sources is influenced by the positive and statistically significant coefficient of access to LPG. The result’s implication is that giving rural households steady access to LPG will increase their use of energy sources for cooking and lighting. Families with consistent access to LPG are more likely to choose LPG as their primary cooking fuel. The results of Ahmad et al. [44], who examined the factors influencing the use of renewable energy sources in Pakistan, also supported this conclusion by finding a positive correlation between the amount of energy sources used and accessibility to them.

Regarding LPG, it is estimated that monthly income received from farming, raising livestock, running a business, and other sources is positive and statistically significant in influencing the number of clean energy sources used by households. This is presumably due to the sense and security of regular income that these revenue streams offer. Additionally, they generate enough cash to cover LPG costs. This finding resonates with the result of Sharma and Dash [45] in their study on household energy use patterns in rural India.

5. Conclusions

This study uses data from rural Nigeria to analyze the determinants of rural households’ choices of cooking and lighting energy sources and their consumption patterns. Evidence from the consumption pattern of rural households reveals that they use multiple energy sources for a particular energy service across different income levels, a phenomenon known as fuel stacking. And this phenomenon is more pronounced in their lighting energy choices than their cooking energy choices. The majority of rural households still use fuelwood for cooking, while few households use modern energy like LPG, with only nine percent reported to have reliable access to LPG. The usage of LPG started from the third income quintile to the fifth income quintile and increased with the income level of households. Contrary to most studies in rural Nigeria, the majority of rural households now use dry cell batteries for lighting, against only 39 percent that uses kerosene, while few also use petrol for lighting, and the usage increases with the income level of households. This may be connected to the limited access to grid electricity in rural Nigeria, as only 23 percent of sampled households reported having reliable access to electricity supply, while 35 percent use electricity for lighting, with 28 percent using rechargeable batteries. Also, 39 percent of households own a personal electricity-generating set, while only 23 percent reported cooking in an indoor kitchen.

More so, results from the application of the multivariate probit model to simultaneously model rural households’ energy choices for cooking and lighting show that household size, access to LPG, education, poverty status, and occupation of household head play an important role in the choice to use clean energy, such as LPG, for cooking, while age of household head, household size, access to grid electricity, possession of electricity-generating sets, education, poverty status of household, occupation of household head, and possession of satellite television play a significant role in the choice to use a clean energy source, such as electricity, for lighting in rural areas. Findings from our study show that an increase in household size decreases the probability of using clean energy such as LPG, but it increases the probability of using fuelwood cooking. Access to LPG increases the probability of using LPG but reduces the probability of using fuelwood, and households that use indoor kitchens are more likely to use LPG and kerosene but less likely to use fuelwood for cooking. An increase in household heads’ education and the education of their spouses reduces the probability of using fuelwood for cooking, while an increase in the education of household heads’ spouses increases the probability of using charcoal, kerosene, and LPG for cooking.

Consequently, the use of quality, affordable, and sustainable energy sources is necessary to lower global environmental pollution. Women are often demonized for providing and using dirty energy in their homes. An improvement in the poverty status of households increases the probability of using LPG and kerosene but reduces the probability of using fuelwood for cooking. Also, households that have access to grid electricity are more likely to use grid electricity and rechargeable batteries but less likely to use kerosene and dry cell batteries for lighting. Households that possess electricity-generating sets are more likely to use petrol and rechargeable batteries but less likely to use grid electricity for lighting. The more educated the household heads, the more the likelihood of using petrol for lighting, while non-poor households are more likely to use grid electricity for lighting. Households that possess satellite television (as compared to those that do not) are more like to use grid electricity and rechargeable batteries but less likely to use dry cell batteries for lighting in rural areas.

In order to encourage environmental sustainability and innovation in the energy sector, legislators are putting sustainable energy research programs into action. Thus, this study recommends new policy changes to enhance rural communities’ access to inexpensive and sustainable energy. In order to achieve highly centralized energy production, authorities ought to encourage off-grid micro-grids. To stimulate the transition of rural households to the use of clean energy for cooking and lighting, the findings from above provide some policy implications on household energy choice in rural areas of Ondo State, Nigeria, and by extension of most developing countries considering the increasing level of multiple fuel usage. Firstly, a reliable and adequate supply of these clean energy sources should be provided to rural areas at affordable prices so as to encourage their usage. Secondly, effective policies and programs that can improve the welfare and poverty status of rural households should be implemented in order to empower them to use clean energy. Thirdly, rural households should be educated on the development of sustainable livelihoods by utilizing contemporary energy services to help low-income households boost their income and productivity. Fourthly, rural households should be provided with the chance to engage in off-farm jobs so as to increase the opportunity costs of using dirty energy sources for both cooking and lighting.